AI-Based Linearization Schemes for 5G/6G Fiber/Wireless Systems

L.A. Melo Pereira, L..L. Mendes, C.J. Albanez Bastos Filho, A. Cerqueira Sodré Junior
National Institute of Telecommunications (Inatel),

Keywords: 5G, 6G, Analog radio-over-fiber, digital pre- distortion, machine learning


We proposed an innovative machine learning-based linearization scheme for analog radio-over-fiber (A-RoF) systems. A centralized radio access network (C-RAN) was employed, which means that a fronthaul link will be used to connect the central office (CO) and the remote radio head (RRH). The fronthaul link will transport data from multiple users by employing A-RoF technology, which means the radio frequency (RF) signal is transported in its analog format, dispensing the use of digital to analog converters (ADCs). On the other hand, analog optical transmissions are more susceptible to linear and non-linear distortions. Linear and non-linear effects will be produced in-band and out-band distortions, leading to reduced signal-to-noise ratio (SNR) and high spurious generation, respectively. We designed the proposed scheme for enhanced remote areas communications (eRAC), which is focused on providing broadband communications for remote and rural areas. Data rates of up to 1 Gbit/s and latencies below 10 ms are expected in this scenario, enabling high-quality video and data acquisition and autonomous and remote-controlled machinery, which demands precision of up to 0.1 m. Up to now, the previous generations of the mobile network have failed in accomplished this task. The high-operational cost to deploy and maintaining complex communications infrastructure is one of the main limitations of this operating scenario. Therefore, cost-effective and low-complex solutions and initiatives are necessary to overcome these limitations. One important initiative is TV white spaces, which aims to use vacant channels of digital television services for broadband communication. In this context, employing C-RAN, A-RoF, and TV white spaces are promising solutions to eRAC communications scenario since they lead to infrastructure simplification and cost reduction. Nonetheless, linear and non-linear effects must be compensated to reduce in-band and out-band distortions. Otherwise, the quality of experience and adjacent leakage ratio will be degraded. Considering the fronthaul link extension to cover remote areas, the aforementioned degradations become even more severe since higher optical and RF are required. Linearization schemes are employed to deal with the non-linear distortions in A-RoF systems. Volterra-based digital pre-distorters and equalizers are widely used in this case. More recently, ML-based solutions are also being employed for linearization and have been demonstrated to be a simpler solution when compared with traditional algorithms with equal or close linearization performance. Nonetheless, ML-based solutions become even more interesting when other effects are also present during signal transmission. For instance, ML solutions can learn complex interactions between the linear and non-linear effects of fiber optics transmissions. Therefore, it can play an even more a distinguished role in the mobile communication system. In summary, artificial intelligence (AI) has significantly affected the telecom market, aiding companies to improve operational efficiency, and user experience and reduce costs. As technology continues evolving, is expected that AI will play an increasingly distinguished role in the telecom industry. For instance, AI will be a crucial component of the sixth generation of mobile network (6G), enabling new applications and capabilities that were not covered by the previous generations.